Paper No 7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Experimental and numerical investigation of biofilm growth and hydrodynamic/biomass
interaction in a granular bioreactor
Sechet Philippe('), Karrabi Mohsen(l), Morra Christophe(l), Cartellier Alain('), Geindreau Christian(3), and
Martins M.F. Jean(2)
1 Laboratoire des Ecoulements Geophysiques et Industriels, UMR 5519 CNRS/UJF/GINP, BP 53X, Cedex 09
Grenoble, France *(Email : philippe.sechet@grenobleinp.fr)
2 Laboratoire des Transferts en Hydrologie et Environnement UMR 5564 CNRS/IRD/UJF/GINP, BP 53X,
Cedex 09 Grenoble, France.
3Laboratoire SolSolideStructureRisque, UMR 5521 CNRS/UJF/GINP, BP 53X, Cedex 09 Grenoble, France.
Keywords: packed bed, coupling hydrodynamicbiomass growth, modelling, experiments
Abstract
The effect of biofilm growth on the clogging of a biofilter is studied on a laboratory scale pilot. At the steady state, pressure,
porosity and biomass concentration profiles are used to study the evolution of the permeability with the biomass content : it
is seen that the EPS and biomass micro structuration have a strong effect on the permeability reduction. That explains why
Kozenylike models found in literature fail to predict the actual permeability values. This finding is of primary importance to
derive correlations or theoretical permeability laws which can be implemented in simple, operational engineering models for
the design of biofilters. Among all the available theoretical models, the Vandevivere model (1995) seems the most
appropriate. In term of biological kinetics, the analysis shows that kinetics parameters are flow dependent which is consistent
with recent observations on the biofilm behaviour grown under different hydrodynamic conditions. One main conclusion is
that the knowledge of macroscopic laws such as the permeabilityporosity relationship is not sufficient to account for the
coupling between hydrodynamic and biomass growth in porous media: those models often use classical kinetic law (Monod,
Haldane...) with constant parameters (at least independent on the flow rate). As biofilms are bacterials communities able to
adapt their metabolisms to external conditions, and in particular to hydrodynamic conditions, it seems that the predictive
character of such models could be improved using flow rate (or shear) dependant effective kinetic parameters to take into
account those effects
Introduction
To optimize the design and operating conditions of biofilters,
it is necessary to have a better understanding of the complex
interactions occurring in these bioreactors. If experimental
or numerical works have been done on the coupling between
the local hydrodynamic and biofilm growth (Picionaru,
1999; Wanner et al, 2004) as well as the coupling between
hydrodynamic and bioclogging in porous media, (Taylor
and Jaffe, 1990, Thullner et al, 2004; Stewart and Kim,
2003; Kapellos et al, 2007) few reliable experiments are
available at pilot scale or industrial biofilter scale on this
problem. Such experiments are however essential to better
understand these systems and help their modeling and
design. This paper aims to present an experiment performed
on a laboratory scale pilot to study the coupling between
biomass growth, hydrodynamic and pollution removal
efficiency in a biofilter. An experimental setup build in
LEGI is presented and preliminary results are shown,
concerning the coupling between permeability reduction and
biomass growth. The experimental results are compared
with existing theoretical work. From literature evidence, it
appears that results obtained at the macroscopic scale may
be not simply explained if the influence of the
hydrodynamic on some structural and metabolic features of
the biofilms is not taken into account. Steady states profiles
are then analysed in the frame of a simple steady state 1D
dimensional model and some more refined features
concerning the coupling between hydrodynamic and biofilm
growth are discussed, in particular a possible biofilm
biological response to the local flow conditions.
Nomenclature
gravitational constant (ms2)
pressure (Nm2)
zlocation in the biofilter
Biomass concentration (g/l)
Carbon source concentration (mg/1)
Oxygen concentration (mg/1)
Biofilm net growth rate (s')
Biofilm detachment rate (s ')
Yield coefficient for oxygen ()
Carbon source saturation constant (mg/1)
Oxygen saturation constant (mg/1)
Carbon source inhibition constant (mg/1)
Superficial velocity (ms1)
Column permeability (m2)
Paper No
Initial permeability
Flow rate (1/h)
Greek letters
0
['max
'Ldecay
Px
P
[1
Porosity ()
Initial porosity
Maximum bacteria growth rate (s')
Bacteria death rate (s')
Biofilm averaged wet density
Water density (kg/m3)
Water viscosity (Pa.s)
Material and Methods
Experimental setup
The experimental setup is mainly based on a transparent
PVC column (1) whose dimension are presented on
figure. The column is filled with expanded clay beads
(Biolite, Degremont). The BioliteO beads are maintained
between two grids to form a granular packed bed and avoid
a possible expansion during the experiment. The beads were
selected in order to get a size distribution as uniform as
possible: their equivalent diameter is about 4.2 mm with an
uniformity coefficient around 1.2. The beads size
distribution can be then considered to be fairly uniform.
Alinentation system
Acquisition system
D=15 cm
H=67 ... .
. .
Nwtvn
(5
Figure 1 Experimental setup overview
The column is fed from an heated constant level water 0.125
m3 tank (2) in order to maintain the water temperature at
300C. Air is bubbling continuously through a diaphragm (4)
in order to dissolve oxygen and maintain a constant oxygen
concentration at the column inlet (6 mg.l'). A nutrient tank
(5) (0.025 m3) contains both phenol and growth nutrient
necessary to the bacterial growth and the biofilm
development. The packed bed is continuously fed with
peristaltic pumps (6,6') (Masterflex) connected to each
tank. Phenol was chosen as the carbon source and the
bacterial strain used was Pseudomonas Putida. In order to
characterize the temporal evolution of biofilm development
inside the column, the biofilter is equipped with two
sampling devices composed of 5 and 9 sampling ports. The
first sampling rack (S1, containing 5 ports evenly
distributed) allows the sampling of the solution inside the
packed bed that is used to monitor quantities such as oxygen
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
or phenol concentrations at different depths in the packed
bed. The second rack (S2) is made up of 9 sampling ports
evenly spacedistributed. These ports are connected to a
multihole pressures tap (7) and a pressure sensor (Druck
LPX9000) (8) allowing the measurement of the relative
pressure along the column with only one pressure sensor.
Standard protocol of manipulation
Column preparation
The standard protocols for the granular media, substrate,
culture medium and bacteria strain preparations as well as
the column feeding can be found in (Karrabi, 2009) and
(Karrabi et al, 2006). Therefore, they won't be detailed here.
For each experiment, the initial permeability Ko of the
packed bed is measured via infiltration experiments
conducted at various flows according to the Darcy's law.
The knowledge of the water volume corresponding to the
column volume filled with beads allowed also the
determination of the initial porosity )o. After physiological
adaptation of the microorganisms to phenol, the biofilter
was submitted to a first stage of colonization by P. Putida,
percolating a culture through the column in a closed loop,
during 24h. After this period of colonization, the system is
switched to the setup used for the biofilm growth
experiments The experiments presented in this paper were
performed at a phenol concentration of 200 ppm at the
column inlet for different total flowrates (5, 10 and 20 1/h)
Pressure, permeability and other parameters measurement
The pressure drop between samplers 1 and 9 (Apl9) is
monitored online with an automat system Field pointTM
associated with LabviewTM software. Every day a complete
pressure profile is measured manually along the column
(Aplj; j=2 to 9).. When the pressure profile is stabilized in
the column (no variation of the mean pressure profile during
3 days), it is considered that a steady state is achieved. The
column is then drained of its solution and emptied, layer by
layer (see Results). The measured pressure drop in each
layer is then used to calculate the layer permeability from
Darcy's equation. When a pressure profile is measured, the
first rack containing the five sampling ports is used to take
effluent samples and measure various quantities such as the
oxygen or phenol concentration profiles.
Results
Existence of a steady state
Figure 2 shows the evolution of the total pressure drop in
the biofilter during the biofilm development. For the 51h
curve presented on that figure, the pressure drop was taken
manually and the continuous data given by the pressure
sensors were not stored. After a lag time, due to the
physiological adaptation of the bacteria to the chemical
conditions, the pressure shows a significant increase. This
characteristic behaviour is linked to bacterial growth and/or
EPS production since the biofilm develops around the
granular clay balls or within the pores and reduces the
Paper No
porosity of the packed bed. As the pressure drop on the
clean column is lower than 0.1 mbar, these values show the
strong influence of the biofilm. A last phase corresponds to
the stabilisation of the mean level of the pressure drop
(respectively 90, 150 and 210 mbars). This state is due to
the equilibrium between different coupled processes such as
the equilibrium between cells division and detachment that
is mainly controlled by the local hydrodynamic conditions
(shear stress), substrate and oxygen availability (growth
inhibition), biofilm structure (thickness,...). Figure 3 shows
typical pressure profiles for Q=20 1/h. The profiles where
calculated from the pressure drop using equation (1). The
pressure was taken to zero at the water level surface and the
pressure drop was neglected to compute the pressure
corresponding to the needle n9.
P, = P + AI ,, + p.g.Az, ,,+ (1)
Contrary to the experimentation of Stewart and Fogler
(2001) on micro models, we don't observe the formation of
a front which, ultimately, propagates through the entire
column. Moreover, in figure 2, the curves shows oscillations
around the mean pressure drop, behaviour also observed by
Stewart and Fogler (2001). These authors explained these
oscillations by different processes which interact. In
particular, the authors interpret the oscillations as both
growth and polymer production combined with
hydrodynamic redistribution of biomass due to biofilm
sloughing and mechanical recapture downstream. The
biofilm sloughing would also create channels within the
plugged part of the porous media, which influence the
nutrient flow and dispersion within the column In our case,
all these observation (figure 2 and 3) means that if such
detachment process exists, in our operating conditions, it
seems that they don't influence the clogging process and
clogging front localisation. This would mean that the
detached biomass fragment are not captured deeper in the
bioniter, and mat, me axial biomass
controlled by local growth.
0 2 4 6 8 10
Day
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
0.8
0.7 20 11h 200 ppm run2 3dTIday
 5tday
o lth day
0.6 7 day
o 13t day
0.5 14th day
cleancolurrn
E 0.4 
0.3 
0.2
0.1 2
0 T
0 50 103 150 230 253 300
pressure (mbar)
Figure 3 Example of pressure profiles
Other experiments, not presented here, were performed on a
longer duration. It was seen (Karrabi, 2009) that, even if the
column behaviour during the unsteady period (the biofilm
maturation) may change from an experiment to the other,
the existence and reproducibility of the steady states was
ensured.
The pressure profiles presented on figure 3 allow also to
precise the notion of steady state, and in particular the
choice to stop the manipulation around the 15th day. It can
be seen that the pressure profiles don't evolve much between
the 10th and 14th day compared to the evolution presented
between the 7th and 14th day. This behaviour was the same
at 101/h and 20 1/h. That's why, for this particular
experiments, the criteria chosen to stop the experimentation
(no mean pressure change for 3 days) seems relatively
justificable.
Biomass content and porosity measurement
repartition is mainly The column is divided into 6 layers delimited by the
sampling ports used for the oxygen and phenol
measurement (see figure 1 Each layer is carefully drained,
at a very low flowrate, through each sampling port. For
some experiments, the collected water volume was used to
compute the mean (mobile) porosity in each layer. When a
layer is drained, the corresponding colonized Biolite is
recovered. The biofilm of each layer is extracted by
vigorous mechanical agitation (200 rpm) in 2 L of distilled
water for 15 minutes. The same washing method is applied
to each Biolite layer, in order to recover the superficially
attached biomass. After beads sedimentation, the
supernatant containing the biomass is collected and stored at
40C before qualitative and quantitative analysis performed
within 24h. The total dry mass is calculated by drying
51hun3 100ml of the samples at 1050C during 24h. This quantity
corresponds to the mass "EPS + microorganism" in the
sample. Resulting porosity and biomass profiles are
12 14 16 18 20 presented on figure 4 and 5.
Figure 2 total pressure drop evolution with time
Paper No
= 0 a.c,
 Q=5 1/h run3 200ppm
 Q=10 I/h run3 200ppm
SQ=20 l/h run2 200ppm
Furthermore, starting from the expression
((z), it can be shown that (Karrabi, 2009) :
of the porosity
0 01 02 03 04 05 06
z (in)
Figure 4 porosity profiles at the steady state
16 Q5 1/h run3 200ppm
1 +o Q=5 l/h run3 200ppm
Q=10 I/h run3 200ppm
 0 Q=20 I/h run2 200ppm
From equation (2) and (3) the slope ac can be related to the
volumetric biomass density px in each layer. The biomass
volumetric density px computed from figure 10 is 25 g/1.
Peyton et al (1995) report value from different authors
ranging from 10 g/1 to 130 g/1. The present value seems thus
coherent.
o, The result above shows that, for our operating conditions 
6 \ high phenol concentration there doesn't seem to have a
2 4 O great change of the biofilm internal structuration along the
2 'o column, at the steady state. This feature is explained by the
fact that the inlet phenol concentration is high, probably far
O
0 above the inhibition concentration. Given the low rate of
0 01 02 03 04 05 06
i 0) apparent degradation obtained from the few measurements
made on the carbon source (results not presented here), this
would mean that in this particular configuration, bacteria
have enough substrate, given the inlet concentration, so that
the bacteria are not starved whatever their location within
Figure 5 biomass profiles at the steady state the column.
From these figures, the relationship between the porosity
and biomass content was studied. Results are shown on
figure 6.
2 0 200
0150
0100
0 050
0 000
500 1000
Biomass concentration (grIl)
Relation permeabilitybiomass concentration
5 /h200ppmrun1 5 Vh200ppmrun2
S1 0 200 ppmrun2 n 10 1A200 ppm run3
20 I 200ppm run1 201 h200ppmrun2
Clement  Thullner Blofilm mode I s 67
ThullnerMicrocolony model s067 Vandevivere le04 f0045
o lh run3 20 ppm
1 00E00
1 0OE01
10 E02
1 ODE03
1 OOE04
1500
Figure 6 relation porositybiomass concentration
The data come from three independent experiments and lead
to a fairly simple relationship between the porosity and the
biofilm volumetric density (taken as the amount of dry
biomass per volume of wet biofilm, according to (Peyton et
al, (1995) definition): in particular, the figure shows that the
biofilm volymetric density dose not depend on the flowrate
1.00E 05 I
000 010 020 030 040 050 OB0 070 080 090 100
relative porosity
Figure 7 relation relative permeability/relative porosity
The pressures profiles and the porosity profiles at the steady
state allowed the derivation of figure 7 which represents the
correlation between the relative column porosity and the
relative permeability in each column slice. Many theoretical
work have been made on the derivation of permeability
models accounting for the bioclogging of porous media.
These models won't be recalled here (see Karrabi ,2009, and
Anthony,2004, for a complete review). However, some of
these models are presented on that figure. The reduction of
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
and on the loading rate, that would confirm the work of
Peyton (1996),at least at high organic load. From Figure 4,
we get:
0 35
03
0 25
' 02 5
015
01
0 05
 ''
o 5 1/h 200 ppm run3
* 10l/h 200ppmrun3
A 20 l/h 200 ppm run2
linear fit
c (z) = PC Z 1
^00 1
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
permeability is very sharp as soon as the biofilm appears, and
this sharp decrease seems related with the existence of
biopolymers (EPS and the existence of a complex
microstructure at the pore scale. The exact role of EPS on the
permeability reduction is still not well understood but their
importance have already be stressed by Shaw et al (1985)
from experiments performed with living or dead bacteria
cells.
Most of the theoretical models (Clement et al, 1996, Thullner
et al, 2002, 2004) are not able to represent the experimental
data. This is partly explained by the biofilm representation in
these models (uniform biofilm on the pore walls, bacterial
colonies, pore throat plugging) which fails to represent the
real complexity of the actual biofilm structure and
distribution. Furthermore most of the assumptions behind
these models consider that the Kozeny law is still valid for
the clogged system. In particular, Clement's model gives
exactly the same behaviour as the Kozeny law. Furthermore,
all these models do not represent properly the effect of the
EPS on the permeability reduction. Vandevivere's model
(1995) seems to better approximate the experimental data.
This model is also based on a given distribution of biomass
within the pore but it allows a sudden transition between a
"continuous biofilm" clogging mode and "pore throat
plugged" mode. This transition is clearly seen on the fitted
curve around 0=0.8. For relatively high porosity reduction,
the permeability seems to reach a "saturation" value. This
saturation corresponds either the biofilm permeability Kmm
(the system is completely clogged and that the water is
actually flowing through the biofilm matrix itself) either to an
effective permeability taking to account the existence of free
space for the fluid inside the porous media : in this case the
shear stress increases along the permeability reduction and
the detached mass of the biofilm keep the pores open (Taylor
and Jaffe,1990). There would exist a steady state
corresponding to a critical shear stress for which the net rate
of biomass removal would balance the net biofilm growth.
Another phenomena, which is not incompatible with the open
pore model, is the existence of channelling inside the
biomass plugs. These channels were observed by Stewart and
al (2001) on their micro model and by the present authors by
visual inspection of the column wall during each
experiments. Furthermore, the maximum permeability
reduction observed in our experiments corresponds to a
minimum macro porosity 0 equal to 0.15 (curve not shown
here). This leads us to support the channelling effect and the
open pore theory although more investigations are needed to
quantify the relative importance of each individual process
depending on the operating conditions.
Biological kinetics
If there are many experimental evidences that the
permeability reduction is closely related to the biomass
distribution at the pore scale, the link between the
permeability reduction and the biofilm structural
characteristic and the biofilm activity themselves are
however not often taken into account and not fully
understood. Figure 9 presents the oxygen profiles measured
at the steady state on the experimental setup.
5
E 4
0)0
S3
8 2
S1
0 0.1 0.2 0.3 0.4
 5 Lh run3
10 Lh run3
e 20 Lh run2
0.5 0.6
Heigth (m)
Figure 9 : oxygen profiles at the steady state
This figure, along figure 4 and 5, show that as the flow rate
increases, the oxygen consumption decreases, the porosity
reduction increase and the biomass concentration
increase within the column : this behaviour seems unusual
as an oxygen consumption decrease would correspond to a
reduction of bacterial activity and biomass production. Data
from figure 4, 5 and 9 were then discussed in the frame of a
simple steady state sectionaveraged ID model. Under the
hypothesis of a local biomass growth, and biofilm
detachment balanced by the net growth, simplified mass
balances for the biomass (eq 4) and oxygen (eq 5) write:
at
8co rx
0a +V(cov)=0 c
at YxO
where c. is the mean biomass concentration (EPS+bacteria)
in a cross section, Co, the mean oxygen concentration at a
given position and v the superficial velocity. In equation
(4), rx is the net biomass production rate and kdet a
function accounting for the detachment rate. In general, kdet
is related to many parameters accounting for the flow (shear
stress z) and the biofilm structure (thickness lb, biomass
concentration c,, biofilm density, mechanical resistance... )
(Kommedal and Blake, 2003). Concerning the growth rate,
no assumption on the exact formulation of r. is introduced
but we expect that function to depend on many parameters
including physicochemical parameters as well as possibly
parameters linked to the hydrodynamic. In equation (5), Yo
is the yield coefficient for the oxygen and all dispersive
terms were neglected. At the steady state, eq (4) and eq (5)
lead to :
rx = kdet (....)
vVco = c
_
From the experimental data presented on figures 4 and 9,
the evolution of the ratio r1Y,0o was computed from eq (7).
To perform this computation, profiles were fitted with
smooth spline function. Only z locations corresponding to
strong gradients on the oxygen profiles were used in order
Paper No
Paper No
to minimize the uncertainty. Figure 10 shows the zaveraged
value of r1Yo versus the flow rate.
3
4 6 8 10 12 14 16 18 20 22
flowrate I/h
Figure 10 : rx/YxO = f(Q)
It is clearly seen that the ratio rY,.o strongly evolves with
the flow rate. The previous analysis cannot be applied to
equation (6) because of the number of unknown parameters
involved in the formulation of kdet. On the other hand, if the
biomass balance must be fulfilled, equation (6) shows also
that the net growth rate must depend on the hydrodynamic
conditions. However, usually, the net growth rate rx is based
on the Monod kinetic law or a related formulation. In our
case, if we take one of the most complete law (namely the
Haldane law in order to take into account possible bacteria
inhibition by phenol), we get:
S max c~(z) C (Z)
k ....) c,(z)2 co(z)+ko
ks + c. (z)+ c7
A decay
where //max, Udecay k k, and ko are the maximum bacteria
growth rate, the biomass mortality rate, the phenol
saturation coefficient, the phenol inhibition coefficient and
the oxygen saturation coefficient respectively. This
formulation does not explicitly involve hydrodynamic effect
such as those due to the shear stress. Moreover, in our case,
given the phenol and the oxygen concentrations, given the
observed phenol reduction in our operating conditions, and
given the range of values found for these kinetics parameter
in the literature (Illuta and Larachi, 2005), this term appears
to be nearly constant (around 0.35 h'). The measured
evolutions with the flow rate cannot be then explained
through this simple model.
Conclusions
In this work, a laboratory scale bioreactor was presented as
well as a data base resulting of controlled experiments
performed on that pilot. The consistency of the data was
carefully checked and the results allowed the study of some
features concerning the biofilm structure within the biofilter.
The raw data showed the existence of a steady state. In
particular, the clogging front reached a stable position,
which was interpreted as an equilibrium between growth
and detachment, the detached fragment beings not
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
recaptured downstream in the porous media. Some features
concerning the bioreactor behaviour under different flow
conditions were deduced from this steady state.
First, the independent measurement of porosity and biomass
concentration (in term of total solid) showed a very simple
relationship between these two quantities allowing the
computation of the volumetric biofilm density. This value is
not very sensitive to the operating conditions. This result
does not allow us however to discuss the evolution of the
biofilm structure with the operating conditions and local
flow conditions. Parameters not measured here can be
involved and the knowledge of more refined data (among
them the evolution of the EPS/bacteria cells ratio, or the
biomass distribution within a section) are required to discuss
this point.
Second, the relationship between the column permeability
and porous media porosity was then checked and the
experimental results were confronted to existing theoretical
models. The reduction of permeability is very sharp as soon
as the biofilm appears, and this sharp decrease seems related
with the existence of biopolymers as well as the existence of
a complex microstructure at the pore scale. This explains
that a Kozeny like formulation fails to predict the
permeability reduction with the biomass content. Advanced
recent models, based on a refined description of the biofilm
and the interaction fluid/biofilm do not capture this sharp
decrease and the predicted permeability reduction is lower
than those observed on the experimentation. The
experimental permeability reduction reaches a saturation
value which can be interpreted following several concept
depending on whether the pores are considered to be
completely filled (close pore theory) or if some pores
remain available to the flow due to an equilibrium between
the biofilm growth and the detachment rate (open pore
theory). From the coupled measurements of permeability
and porosity reduction, our experiments would favour the
open pore theory (with possibly, the existence of
channelling inside the plug part of the column, as inferred
by Stewart and Fogler (2001) in their microscale
experiment).
Different permeability models were tested on the data. The
Vandevivere model is the most appropriate in the case of our
experiment. That result is important for some models such
as Kildgaards et al (2001) and Brovelli et al (2'i "'J which
use directly a constitutive law between permeability and
porosity to perform the coupling between the momentum
equation and biomass growth. As for the volumetric density,
the permeabilityporosity relationship is not sensitive to the
operating conditions in our case. The sensitivity of this law
to the biofilm structure at the pore scale as well as the
biomass distribution at the pore scale is still an open
question and required more data on a larger range of
operating conditions.
Analysing oxygen profiles along the porosity and dry matter
concentration, it appeared that higher flow rate led to lower
oxygen consumption (so seemingly less active biomass) but
led also to a higher amount of biomass inside the column.
The steady state profiles were thus analysed in the frame of
a simple ID model. It was shown that classical kinetics law
Paper No
accounting for cells growth (Monod, Haldane laws) or
biofilm effective growth rate (through the use of a limiting
function) could not explain our experimental data, if the
biomass growth rate is not flow rate (or shear) dependant.
This result could be interpreted as a higher polymer
production as the flow rate increases, which is coherent with
reported biofilm behaviour with increasing shear stress.
As a general conclusion, we showed that simple global
measurements made at the global scale gave results which
can be interpreted physically in term of biofilm behaviour.
Some of our conclusions are consistent with previous
finding in other context (such as Loodsrecht et al, 2002,
results in the case of an airlift reactor). A contrario, global
experiments must also be supplemented with information
gathered on local scale experiments. In particular, the
modelling of the permeability reduction with the biomass
content is still an open problem: our results show that
models based on a modification of the Kozeny law or
simple geometrical assumptions on the biomass distribution
at the pore scale can not reproduce the data in our operating
conditions. The biofilm structure and composition (in
particular the EPS content) seems to play a great role, the
mechanisms involving the EPS being not completely
elucidated.
Finally, the results show also that the knowledge of the
permeability reduction law, along a rough description of the
biofilm is not enough to predict the porous media
bioclogging. Our results at the global scale show an effect of
the flow on the biological kinetics which could be linked to
the bacteria biological response to the local hydrodynamic.
This finding is consistent with recent results about biofilm
behaviour (EPS production: Qi et al, 2008, growth kinetics:
Tsai, 2005, biofilm phenotype: Simoes et al, 2007). The
effect of the flow on the various kinetic laws accounting for
the effective biofilm components growth rate is then crucial.
To get realistic models, describing separately the EPS and
the bacterial cells seems necessary, as well as including
phenomena linked to the coupling between the cells
metabolism and the flow (growth rate, EPS production,
gene activation ...).
Acknowledgements
The authors would like to thanks the CNRS, for the funding
of this research through the ACI program
References
Anthony P. Modelisation des couplages hydrodynamique
biomasse en milieu poreux, (Modeling of coupling between
hydrodynamic and biomass in porous media), Master thesis,
3SR Laboratory, University of Grenoble, France (in french)
02'"' 4)
Clement, T.P., Hooker, B.S. and Skeen, R.S., Macroscopic
models for predicting changes in saturated porous media
properties caused by microbial growth, Ground Water, 34,
pp 934942 (1996)
7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
Kapellos, G. E., Alexiou, T. S. and Payatakes, A. C., A
multiscale theoretical model for diffusive mass transfer in
cellular biological media, Mathematical Biosciences, 210,
pp 177237 (21 ')
Karrabi M., Coupling between hydrodynamic and biofilm in
porous media : application to a biofilter. Ph D Thesis, LEGI,
University of Grenoble, France (2'" "'
Karrabi M., Sechet P, Morra Ch., Florensa I, Geindreau Ch,
Martins J, and Cartellier A., Experimental investigation of
biofilm growth and hydrodynamic/biomass interaction in a
granular bioreactor, proceedings of the 6th Chisa Congress,
Pranha, Czech Republic (2006)
Peyton, B.M., Skeen, R.S., Hooker, B.S., Lundman, R.W.
and Cunningham, A.B., Evaluation of bacterial detachment
rates in porous media., Appl. Biochem. Biotechnol. 51, pp
785797 (1995)
Picioreanu, C., Multidimensional modeling of biofilm
structure, Ph.D. thesis, Delft University of Technology, The
Netherlands. (1999)
Qi P., Wang W. and QI Z., Effect of Shear Stress on
Biofilm Morphological Characteristics and the Secretion of
Extracellular Polymeric Substances, 978142441748
3/08/2008 IEEE (2' ',)
Shaw, J. C. and Bramhill, B. and Wardlaw, N. C. and
Costerton, J. W., Bacterial Fouling in a Model Core System,
Applied and Environmental Microbiology, 49, pp 693701
(1985)
Simoes M., Peireira M.O., Sillankorva S., Azeredo J. and
Vieira M.J., The effect of hydrodynamic conditions on the
phenotype of Pseudomonas fluorescens biofilms,
Biofouling, 23 (3/4), pp 249258 (2 '"')
Stewart, T. L. and Fogler, H. S., Biomass plug development
and propagation in porous media, Biotechnology and
Bioengineering, 72(3), pp 353363 (2001)
Stewart, T. L. and Kim, D. S., Modeling of biomassplug
development and propagation in porous media, Biochemical
Engineering Journal, 17, pp 107119 (2" 11 ')
Taylor, S.W. and Jaffe, P. R., Biofilm growth and the
related changes in the physical properties of a porous
medium. 1. Experimental investigations", Water Resource
Research, 26, pp 21532159 (1990)
Thullner, M. ,Mauclaire, L. Schroth, M. H.,Kinzelbach, W.
and Zeyer J., Interaction between water flow and spatial
distribution of microbial growth in a twodimensional flow
field in saturated porous media, Contaminant Hydrology,
58, pp 169189 ('1 "i2)
Thullner, M., Schroth, M. H., Zeyer, J. and Kinzelbach, W.
, Modeling of a microbial growth experiment with
Paper No 7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30June 4, 2010
bioclogging in a twodimensional saturated porous media
flow field, Contaminant Hydrology, 70, pp 3762 (2I" i4)
Tsai YP. Impact of flow velocity on the dynamic behaviour
of biofilm bacteria, Biofouling 21 (5/6), pp 267277 (2005)
Vandevivere, P., Bacterial clogging of porous media: a new
modeling approach, Biofouling, 8, pp 281291 (1995)
Wanner, O., Eberl, H. J., Morgenroth, E., Noguera, D. R.,
Picioreanu, C. Rittmann, B. E. and Van Loosdrecht, M.
C.M., Mathematical Modeling of Biofilms, IWA Biofilm
Specialists Conference in Las Vegas, Nevada, USA (2i" 4)
